Structural equation modelling using Mplus

Description:

Summary

This is a hands-on course intended for applied doctoral or post-doctoral researchers who are familiar with regression methods and who want to broaden and deepen their range of analytical skills. The full course comprises four days; the first two days introduce the building blocks of Structural Equation Modelling (SEM). The second two days cover Mplus’ advanced extensions to SEM, which put unprecedented modelling power and flexibility into the hands of the applied researcher.

Part 1 (days 1-2)

SEM extends standard regression modelling in two fundamental ways. First, SEMs can contain a Mediator/Path/Structural element, whereby multiple outcome variables can be included in the same model, and where these variables can be arranged in ‘causal’ chains. Second, SEMs can contain unobserved, latent predictors, e.g. latent factors.

Part 2 (days 3-4)

Part II of the course shows how greater modelling flexibility can be gained by exploiting Mplus’ advanced features for dealing with non-normally distributed variables. Students will be shown how to extend factor analysis to binary and ordinal outcome variables. The assumption that latent variables should be normally distributed can also be relaxed, allowing for latent groups or classes of individuals to be modelled. We also consider the topic of measurement invariance across groups, i.e. the assumption that the latent variables actually ‘mean’ the same things to different people.

The course will:

1. Introduce the student to a suite of Structural Equation and Latent Variable models
2. Show how to estimate these models using the Mplus statistical package.
3. Show how to interpret the output of these models.
4. Show how these models can be used to address substantive research questions.

Course tutors

The course will be taught by Dr Nick Shyrane and Dr Bram Vanhoutte

Course requirements 

We assume that students are experienced users of regression models (linear and logistic). No prior experience of using Mplus is required.

If you are an experienced user of normal-theory SEMs and are looking to learn about the advanced features in Mplus for dealing with non-normality, it may be possible to attend just Part 2 of the course; please contact the course convenor (nick.shryane@manchester.ac.uk) to discuss. 

Useful Reading 

Part 1: Byrne, B. M. (2012). Structural Equation Modeling with Mplus. Basic Concepts, Application, Programming. New York, NY: Routledge.

Part 2: Bartholomew, Knott & Moustaki (2011). Latent Variable Models and Factor Analysis: A Unified Approach (3rd Ed.). London: Wiley. 
Muthen, B. O. (2002). Beyond SEM: General latent variable modelling. Behaviormetrika, 29, 81-117.

Cost:

Students £600 | University of Manchester staff £600 | other attendees £900

Website and registration:

http://www.methods.manchester.ac.uk/events/summer-school-2016/

Region:

North West

Keywords:

Quantitative Data Handling and Data Analysis, Regression Methods, Linear regression, Log-linear regression, Logistic regression, Multilevel Modelling , Latent Variable Models, Structural equation models, Quantitative Software, Mplus

Related publications and presentations:

Quantitative Data Handling and Data Analysis
Regression Methods
Linear regression
Log-linear regression
Logistic regression
Multilevel Modelling
Latent Variable Models
Structural equation models
Quantitative Software
Mplus

Back to archive...